Dynamics Analysis Of Magnetically Controlled Memristor Synaptic Coupled Hopfield Neural Networ

Because of its memory and transmission characteristics,memristor is considered as the most ideal device to replace the nerve synapse.Hopfield neural network(HNN)model is a neural network model created based on the biological neural structure,which can better simulate the nonlinear dynamical behaviors of the brain.The traditional HNN model is weakly nonlinear and does not take into account the influence of synapses on the information transmission between neurons.In order to better simulate the complex electrical signal behavior of the brain,researchers have proposed the memristor synapsecoupled HNN model.In the memristor synapse-coupled HNN model,due to the time-varying external stimulation received by biological neurons,the electrical response speed inside neurons(neuron activation gradient)and the connection strength between neurons(synaptic weight)will vary with the change of external stimulation.Therefore,it is necessary to study the influence of the change of neuron activation gradient and synaptic weight on the dynamical behavior of the memristor synapse-coupled HNN model.In this thesis,an ideal flux-controlled memristor synapse was used to simulate the induced current generated by the potential difference between two neurons,and a memristor synapsecoupled dual-neuron HNN model with adjustable neuron activation gradient and synaptic weight(hereinafter referred to as the derived memristor HNN model)was proposed.The effects of neuron activation gradient and synaptic weight on the stability and dynamical behavior of the derived memristor HNN model was analyzed.Matlab is used to conduct numerical simulation on the derived memristor HNN model of the two-parameter bifurcation diagram,dynamical map,local basin of attraction and phase diagram and other analysis methods,from which the following complex phenomena are found: the unbounded region alternates with the stable point region;the unbounded region has the stable point boundary;the values of the neuron activation gradient and synaptic weight together affect the morphology of the derived memristor HNN model;coexistence of multistable states such as stable point attractors,period-1oscillations,chaos and unbounded states;the derived memristor HNN model contains hidden phenomena such as transient cycles,intermittent chaos,coexisting transient chaos and transient proposed cycles.The corresponding circuit was designed based on the derived memristor HNN model,and the circuit was simulated by using Multisim.The hardware circuit was built according to the simulated circuit,and it was verified that the derived memristor HNN model will appear period-1,period-2,period-4 and chaos states with the change of parameters.The experimental results are in good agreement with the theoretical analysis.